Stability AI finally addressed its absurd licensing. Will it be enough to regain trust from the community?
Stability AI, the company behind the popular open-source image model Stable Diffusion, has responded to the community’s concerns about its licensing terms.
Three weeks ago, they released the Stable Diffusion 3 model with 2 billion parameters (SD3 Medium), which faced criticism for its confusing license and obvious performance issues.
Even the popular AI model-sharing platform Civit AI has temporarily banned SD3 Medium while clearing matters with Stability AI.
Today, Stability AI released an update to the licensing to address these concerns.
We acknowledge that our latest release, SD3 Medium, didn’t meet our community’s high expectations. We heard you and have made improvements to address your concerns and to continue to support the open source community.
The updated license is now more permissive, expanding the scope of permitted use significantly compared to its predecessor. Key changes include:
These changes reflect a shift towards a more comprehensive and flexible licensing approach, balancing open access with commercial considerations.
You can read more details about what’s changed in the licensing terms in the official blog post.
Scope of Use
Commercial Use
Registration Requirement
Revenue Threshold
Intellectual Property
In addition to licensing issues, Stability AI acknowledges the poor quality of SD3 2B. Users noticed significant degradation in quality, such as prompts where the subject lies on the grass.
Before we released SD3 Medium, our initial testing indicated that it was, in most cases, a much better base model compared to SDXL, in terms of prompt adherence, diversity, detail, and overall quality. However, the community quickly identified some critical quality issues mainly related to body poses and words that were too rarely seen in the training set.
Stability AI is now working to improve these key areas.
Despite the clearer and more positive licensing, the problem with degrade in performance remains. A new version of SD3 Medium is already in the works. Hopefully, when the updated model comes out, this whole fiasco is still salvageable.
Also, in the past several weeks Stability AI has already burned a bunch of goodwill in the community. Will the clarifications and acknowledgment of the poor quality be enough to bring back the trust?
For now, it’s best to remain skeptical and wait for the release of the updated model. I expect the 8B model to be way better than the 2B.
Software engineer, writer, solopreneur